Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521https://doi.org/10.1186/s12889-019-6775-4

RESEARCH Open Access

Population-based bloodstream infection

surveillance in rural Thailand, 2007–2014 Julia Rhodes1*, Possawat Jorakate1, Sirirat Makprasert1, Ornuma Sangwichian1, Anek Kaewpan1,Thantapat Akarachotpong1, Prasong Srisaengchai1, Somsak Thamthitiwat1, Supphachoke Khemla3,Somkid Yuenprakhon4, Wantana Paveenkittiporn5, Anusak Kerdsin5,6, Toni Whistler1,2, Henry C. Baggett1,2

and Christopher J. Gregory1,2

Abstract

Background: Bloodstream infection (BSI) surveillance is essential to characterize the public health threat ofbacteremia. We summarize BSI epidemiology in rural Thailand over an eight year period.

Methods: Population-based surveillance captured clinically indicated blood cultures and associated antimicrobialsusceptibility results performed in all 20 hospitals in Nakhon Phanom (NP) and Sa Kaeo (SK) provinces. BSIs wereclassified as community-onset (CO) when positive cultures were obtained ≤2 days after hospital admission andhospital-onset (HO) thereafter. Hospitalization denominator data were available for incidence estimates for2009–2014.

Results: From 2007 to 2014 a total of 11,166 BSIs were identified from 134,441 blood cultures. Annual CO BSIincidence ranged between 89.2 and 123.5 cases per 100,000 persons in SK and NP until 2011. Afterwards, COincidence remained stable in SK and increased in NP, reaching 155.7 in 2013. Increases in CO BSI incidenceover time were limited to persons aged ≥50 years. Ten pathogens, in rank order, accounted for > 65% of COBSIs in both provinces, all age-groups, and all years: Escherichia coli, Klebsiella pneumoniae, Burkholderiapseudomallei, Staphylococcus aureus, Salmonella non-typhi spp., Streptococcus pneumoniae, Acinetobacter spp.,Streptococcus agalactiae, Streptococcus pyogenes, Pseudomonas aeruginosa. HO BSI incidence increased in NPfrom 0.58 cases per 1000 hospitalizations in 2009 to 0.91 in 2014, but were higher (ranging from 1.9 to 2.3)in SK throughout the study period. Extended-spectrum beta-lactamase production among E. coli isolates andmulti-drug resistance among Acinetobacter spp. isolates was common (> 25% of isolates), especially among HOcases (> 50% of isolates), and became more common over time, while methicillin-resistance among S. aureusisolates (10%) showed no clear trend. Carbapenem-resistant Enterobacteriaceae were documented in 2011–2014.

Conclusions: Population-based surveillance documented CO BSI incidence estimates higher than previously reportedfrom Thailand and the region, with temporal increases seen in older populations. The most commonly observedpathogens including resistance profiles were similar to leading pathogens and resistance profiles worldwide, thus;prevention strategies with demonstrated success elsewhere may prove effective in Thailand.

Keywords: Bloodstream infections, Community-acquired infections, Healthcare-associated infections, Antimicrobialresistance, Population-based surveillance, Thailand, Global health security

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

* Correspondence: icq0@cdc.gov1Global Disease Detection Center, Thailand Ministry of Public Health (MOPH)– United States Centers for Disease Control and Prevention (CDC)Collaboration, Nonthaburi, ThailandFull list of author information is available at the end of the article

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 2 of 12

BackgroundBloodstream infections (BSI) are important causes ofmorbidity and mortality with incidence rates compar-able to stroke, acute myocardial infarction and trauma[1] but the public health threat of BSI has beenpoorly characterized in Southeast Asia. BSI burdenestimates are often limited to North America andEurope due to a lack of population-based studies inother parts of the world [2]. Population-based BSI in-cidence estimates from Southeast Asia are especiallyscarce, but pathogen-specific studies and reports ofantimicrobial resistant pathogens indicate a substan-tial BSI burden in Southeast Asia [3–7]. A systematicreview of community-acquired BSIs in south andSoutheast Asia, found that 9% of hospitalized, febrilepatients with blood cultures collected had bacteremiaand 9% of these patients died in-hospital [8].While this information is useful, additional descrip-

tions of BSI epidemiology from South East Asia areneeded to characterize the public health threat by: 1)allowing for comparison between BSI disease burdenand trends versus other health conditions, 2) assessingthe relative importance of specific BSI pathogens, and 3)identifying high-risk sub-populations for development ofeffective treatment regimens and targeting of publichealth prevention interventions. Estimation of the over-all BSI burden is particularly useful as prevention strat-egies are not necessarily pathogen specific and overallestimates are needed to evaluate the potential impact ofinfection control and antimicrobial resistance programs.Furthermore, characterization of BSI epidemiology facili-tates rapid detection and containment of public healththreats at their source, thereby enhancing global healthsecurity.From 2005 through 2014, the Thailand Ministry of

Public Health - US Centers for Disease Control and Pre-vention (CDC) Collaboration, together with the NakhonPhanom and Sa Kaeo Provincial Health Offices, con-ducted population-based BSI surveillance in Sa Kaeoand Nakhon Phanom provinces. In this paper wesummarize the overall findings.

MethodsSettingSa Kaeo is located in eastern Thailand near the Cambo-dian border; Nakhon Phanom is in northeast Thailandnear the Laos border. These two rural provinces have acombined population of 1.1 million. All 20 hospitals inthe two provinces participated, including two provincialhospitals (225–327 bed referral centers) and 18 periph-eral, district hospitals (10–140 beds). There were no pri-vate hospitals in either province and Thai citizens areprovided with health care at minimal or no cost.

Specimen collection and processingBlood cultures were performed as clinically indicated perphysician request among hospitalized patients. From2007 to 2010, we encouraged blood cultures for hospital-ized patients with suspected pneumonia and for patientsaged ≤5 years old with possible sepsis, by reimbursinghospitals for the costs of culture for these patients.Nurses, regularly trained in phlebotomy, collected speci-mens from a single peripheral site. According to proto-col targets, specimens from district hospitals weretransported at 15–30 °C within 24 h of collection to pro-vincial hospital laboratories, where all specimens wereprocessed using an automated blood culture system(BacT/ALERT® 3D, bioMérieux, U.S.A.). Nurses com-pleted a form with patient details including antibioticuse in the last 72 h.Each blood specimen was divided into a bottle opti-

mized for aerobic growth [BacT/ALERT FAN Aerobic(FA) for ages ≥ 5 years (target volume of 10 mL) andBacT/ALERT Pediatric FAN (PF) for ages < 5 years (tar-get volume of 4 mL)] and a bottle for enhanced growthof mycobacteria, fungal pathogens, and other fastidiousagents [BacT/ALERT Mycobacteria Blood (MB) (targetvolume of 3 mL)]. If blood volume was insufficient to in-oculate both bottles at targeted levels, the FA/PF bottlewas prioritized. The protocol was modified in October2011; thereafter, MB bottles were not routinely proc-essed, but available by physician request, (e.g. for casesof suspected tuberculosis or Burkholderia pseudomalleiinfection) [9]. FA/PF and MB bottles were incubated inthe BacT/ALERT system at 35 °C for up to 5 and 42 daysrespectively, or until the instrument signaled a positive re-sult for growth (i.e., alarm positive). Media from alarmpositive bottles was sub-cultured onto sheep blood, choc-olate and MacConkey agar plates and incubated overnightat 35 °C. Standard biochemical testing was used for identi-fication [10]. From 2012, API® identification strips (bio-Mérieux, U.S.A.) were used when standard testing was notdefinitive. Confirmatory identification of all pathogenswas performed at Thailand’s National Institute of Healththrough the end of 2010, when confirmatory testing waslimited to selected species (e.g. Streptococcus spp., B. pseu-domallei, Salmonella non-typhi spp., Vibrio spp. and un-usual pathogens). Confirmatory testing was discontinuedin 2012; however participation in national External Qual-ity Assurance (EQA) programs has been ongoing through-out the surveillance period and an international EQAprogram (Royal College of Pathologists of Australia Qual-ity Assurance Programs Pty. Ltd., Australia) was initiatedin 2011 and continued throughout the remainder of thesurveillance period.Antimicrobial susceptibility testing for 1) extended

spectrum beta-lactamase (ESBL) E. coli and K. pneu-moniae, 2) methicillin-resistant S. aureus (MRSA), 3)

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 3 of 12

vancomycin-resistant S. aureus (VRSA), 4) penicillin-resistant S. pneumoniae, and 5) carbapenem-resistantenterobacteriaceae (CRE) was performed according tothe Clinical and Laboratory Standards Institute guide-lines [11]. ESBL screening test results were used todetermine ESBL status when ESBL E. coli and K.pneumoniae confirmatory testing was not available.This definition was based on prior work which foundscreening test sensitivity > 95%, specificity > 90%,and > 90% agreement with confirmatory tests for bothE. coli and K. pneumoniae [12, 13]. All CRE caseswere confirmed by retesting at the Thailand NIH. Aci-netobacter spp. isolates were defined as multi-drug re-sistant (MDR) if they were resistant to 3 or more ofthe following drug classes by disc diffusion testing:aminoglycosides (Amikacin or Gentamycin); cephalo-sporins (Cefotaxime or Ceftazidime or Cefoperazone);fluoroquinolones (Ciprofloxacin); carbapenems (Imipe-nem or Meropenem) [14].Serotyping for S. pneumoniae was performed using

multiplex PCR [15]. For isolates that could not be typedby this method, Quellung serotyping was done at theStreptococcus Reference Laboratory, U.S. CDC in At-lanta, Georgia. Serotyping for Haemophilus influenzawas performed using real-time, multiplex PCR [16] atthe Thailand National Institute of Health.

Definitions and population dataBacterial isolates were defined as likely pathogens if atleast one culture bottle (FA/PF or MB) grew an organ-ism likely responsible for infection [17] and did not growa likely contaminant in the same bottle; S. pneumoniae,B. pseudomallei and Salmonella non-typhi spp. wereconsidered pathogens regardless of other isolategrowth. Common skin and environmental organismswere considered likely contaminants (Fig. 1). Forpatients with multiple positive cultures, the first posi-tive culture to grow a pathogen was included as aBSI case. Repeat positive cultures that grew the samespecies within 30 days were excluded with the excep-tion of B. pseudomallei for which cases were onlycounted once regardless of the timing of subsequentpositive cultures.BSI cases were classified as hospital-onset (HO) or

community-onset (CO) based on definitions adaptedfrom MRSA surveillance [18]; HO cases were defined asthose with pathogens from blood specimens collected >2 days after the hospital admission date and CO caseswere those collected ≤2 calendar days after admission.CO incidence rate denominators were derived from

Sa Kaeo and Nakhon Phanom provincial populationprojections for 2010–2014 from the 2010 NationalEconomic and Social Development Board (NESDB)of Thailand [19]. For the period 2007–2009,

age-stratified official intercensal estimates were notavailable. Instead, NESDB provided revised overallprovincial population estimates for 2007–2009 basedon the 2010 census. To derive age-specific populationestimates we applied the 2010 NESDB age distribu-tion to 2007–2009 NESDB overall provincial popula-tion estimates [20].HO incidence rate denominators were calculated using

the number of hospitalizations from 2009 to 2014 asfound in hospital administrative databases (Naorat, S., Pir-alam, B., personal communication, 2015). In Sa Kaeo, fourdistrict hospitals were missing 2009 data. Together, thesehospitals account for ≤15% of annual hospitalizations inour surveillance system. After careful examination ofhospitalization trends, we imputed 2009 hospitalizationdata with observed 2010 data for these four hospitals.Military hospitals in Sa Kaeo and Nakhon Phanom prov-inces were also excluded from HO incidence calculations,as hospitalization data was not available for Nakhon Pha-nom and only available for two years in Sa Kaeo (2013–2014). Military hospitals comprised < 1% of all hospitaliza-tions in each year with available data.Annual incidence estimates were calculated as the

number of cases in each year divided by the popula-tion estimate for each year and reported as cases per100,000 persons. Ninety-five percent confidence inter-vals (CIs) on incidence estimates were calculatedbased on a Poisson distribution using the exactmethod. The statistical significance of trends overtime in incidence rates was calculated by fitting a lin-ear regression model with ‘year’ as the only predictorof annual incidence estimates. P-values <.05 for theyear variable coefficient indicated significant trendsover time. Analyses were conducted using SAS ver-sion 9.3 (SAS Institute Inc., Cary, NC, U.S.A.)

ResultsFrom 2007 to 2014 a total of 134,441 blood cultureswere processed for 128,503 patients. The annual numberof patients cultured increased from 15,727 in 2007 to17,893 in 2014. Blood volume targets were met for FA/PF bottles for 93% of cultures among persons ≥5 yearsold and 28% of children < 5 years old. District hospitalsaccounted for 45% of blood cultures and 44% of patho-gen positive cases, with the remainder coming from pro-vincial hospitals. Specimens from district hospitals wereplaced in the BacT/ALERT instrument within 24 h ofcollection for 89% of cultures. Overall, 16% of processedcultures signaled positive: a pathogen was isolated from9.0%, 5.0% grew a likely contaminant only, 1.3% werealarm positive, but had no growth on sub-culture andthe remainder were not identified (Fig. 1). Cultures posi-tive for only contaminants ranged from 4.1% in 2014 to6.1% in 2012.

Fig. 1 Flow diagram of blood cultures results, Sa Kaeo and Nakhon Phanom provinces, Thailand, 2007–2014

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 4 of 12

Of 11,166 BSIs, 10,007 were CO, 1109 were HO, and50 could not be determined as CO vs HO (Fig. 1). Inprovincial hospitals, equal proportions of CO and HOcases came from surgical wards (16%) and fromnon-surgical, in-patient wards (61%); however, theremaining CO cases came from emergency rooms (16%)and intensive care units (ICUs) (6%), while all ofremaining HO cases came from ICUs (24%), (ward dataavailable for provincial hospitals from 2010 to 2014only).From 2007 to 2014, overall CO BSI incidence was 110

(95% CI: 98, 123) cases per 100,000 population with asignificant increase over time (Table 1). Overall HO BSIincidence was 1.3 (95% CI: 0.9, 1.7) cases per 1000 hos-pitalizations with no significant trend. CO BSI incidencewas highest for persons 65+ years, followed by those age50–64 years, and children < 5 years old. HO BSI inci-dence was also highest among older persons but did notdiffer between 50–64 year olds and 65+ year olds. In-creases in CO BSI incidence over time were limited topersons aged 50 years and older. Increases in incidence,together with increasing populations in older age groups,translated into increases in case counts: from 263 casesin 2007 to 492 cases in 2014 among 50–64 year olds(46% increase); from 343 cases in 2007 to 572 cases in2014 among those 65 years and older (40% increase).

CO BSI incidence (95% CI) was similar in Sa Kaeo andNakhon Phanom and ranged between 89 (81, 97) and124 (114, 133) cases per 100,000 population until 2011(Fig. 2a). Afterwards, incidence increased to a high of156 (145, 166) in 2013 in Nakhon Phanom (p-value fortrend < 0.01), while no clear increase was observed in SaKaeo. The higher 2013–2014 incidence in Nakhon Pha-nom corresponded with increases in several pathogens:E. coli (45% increase), K. pneumoniae (33%), S. aureus(36%), S. pneumoniae (38%), and Pseudomonas aerugi-nosa (32%).HO BSI incidence was higher in Sa Kaeo province

compared to Nakhon Phanom (Fig. 2b). In Sa Kaeo, HOBSI incidence ranged from 1.9 to 2.3 cases per 1000 hos-pitalizations and did not increase significantly over time;whereas, in Nakhon Phanom HO incidence (95% CI) in-creased from 0.58 (0.40, 0.77) cases per 1000 hospitaliza-tions in 2007 to 0.91 (0.68, 1.14) in 2014, p < 0.01.Provincial (i.e. referral) hospitals had HO BSI incidences2.7–6.6 times higher than district level hospitals (datanot shown).Gram-negative bacteria accounted for 73% of CO

BSI, Gram-positive bacteria 25%, and fungi 2.2% (1.6%Cryptococcus neoformans (n = 165); 0.6% Candida spp.(n = 56)). CO Gram-negative BSIs peaked during Julyto October largely due to seasonal increases in B.

Table

1Ann

ualcom

mun

ity-onset

(CO)andho

spital-o

nset

(HO)aBSIinciden

ceb,cby

agein

SaKaeo

andNakho

nPh

anom

provincesThailand

,2007-2014

Age

(yrs.)

Ann

ualB

SIIncide

nce(95%

CI)

p-value

fortren

dOverall

incide

nce

(95%

CI)

2007

2008

2009

2010

2011

2012

2013

2014

Overall

CO(N=10,007)

91(80,103)

91(79,102)

111(99,123)

118(105,131)

100(88,111)

108(96,120)

133(119,146)

130(117,143)

<.05

110(98,123)

HO(N=917)

1.1(0.71,1.5)

1.2(0.8,1.6)

1.4(0.9,1.8)

1.5(1.0,1.9)

1.1(0.8,1.5)

1.5(1.0,1.9)

NS

1.3(0.9,1.7)

<5 CO(N=522)

107(86,115)

82(63,90)

57(41,64)

77(59,84)

46(32,52)

84(65,90)

71(53,76)

71(53,76)

NS

75(57,93)

HO(N=107)

0.6(0.3,0.9)

0.7(0.3,1.0)

0.8(0.5,1.2)

0.9(0.5,1.2)

0.6(0.3,0.9)

0.7(0.4,1.1)

NS

0.7(0.4,1.0)

5-14 CO(N=187)

15(10,42)

15(9.6,41)

9.5(5.2,28)

13(7.7,35)

10(5.7,29)

7.2(3.5,21)

13(7.6,34)

13(7.7,34)

NS

12(7.1,17)

HOd(N=11)

NA

NA

NA

NA

NA

-0.2(0.0,0.5)

15-49

CO(N=2,739)

56(50,62)

59(52,65)

72(65,79)

70(63,77)

60(53,66)

53(47,59)

66(60,73)

64(57,70)

NS

62(56,69)

HO(N=234)

0.9(0.6,1.2)

1.0(0.7,1.3)

1.0(0.7,1.3)

0.8(0.5,1.0)

0.7(0.5,1.0)

0.8(0.5,1.0)

NS

0.9(0.6,1.1)

50-64

CO(N=3,019)

137(120,153)

144(127,161)

193(173,213)

203(182,223)

176(157,194)

184(165,203)

247(225,268)

225(205,245)

<.01

190(171,209)

HO(N=262)

2.1(1.4,2.7)

2.4(1.7,3.1)

1.9(1.3,2.5)

2.7(2.0,3.4)

1.7(1.1,2.2)

3.0(2.2,3.7)

NS

2.3(1.6,3.0)

65+ CO(N=3,540)

342(306,378)

328(292,364)

424(383,464)

466(423,509)

385(347,423)

448(407,488)

500(458,541)

501(460,542)

<.01

426(387,466)

HO(N=303)

1.7(1.2,2.3)

1.9(1.3,2.5)

2.5(1.8,3.2)

2.8(2.1,3.5)

2.4(1.8,3.1)

2.6(1.9,3.2)

.06

2.3(1.7,3.0)

a Den

ominator

data

forHOincide

ncecalculations

availableafter20

08.

bCOincide

nce:

cases/10

0,00

0po

pulatio

nc HOincide

nce:

cases/1,00

0ho

spita

lizations

dAnn

ualinciden

ceestim

ates

notshow

nas

yearly

case

coun

tswere<5

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 5 of 12

a

b

Fig. 2 a/b Annual community-onset (CO) and hospital-onset (HO) BSI incidence in Sa Kaeo and Nakhon Phanom provinces,Thailand, 2007–2014

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 6 of 12

pseudomallei. There was little difference in COGram-positive BSI frequency throughout the year(Additional file 1: Figure S1).Ten pathogens accounted for more than two-thirds

of CO BSIs in both provinces, in all age-groups (exceptnewborns < 1month old), and in all years: E. coli, K. pneu-moniae, B. pseudomallei, S. aureus, Salmonella non-typhispp., S. pneumoniae, Acinetobacter spp., S. agalactiae, S.pyogenes, and P. aeruginosa. Polymicrobial infections wereobserved in 4.1% of cases with E. coli and K. pneumoniaeco-infections being the most common. Pathogen

distributions for patients aged 15–49, 50–64, and 65 yearsand older were similar with E. coli as the most commonpathogen, followed by B. pseudomallei and K. pneumoniae(Fig. 3), though E. coli accounted for an increasing propor-tion of cases with age. C. neoformans accounted for 4.7%of cases among the 15–49 year olds and less than 2%among all other age groups. Pathogen distributions variedsubstantially among children. S. aureus accounted for 31%and 30% of CO BSIs among newborns and 5–14 year-olds,respectively, but only 6% among 1–4 year olds. Instead,two potentially vaccine preventable pathogens, S.

Fig. 3 Ten most commonly observed community-onset (CO) BSI pathogens by age in Sa Kaeo and Nakhon Phanom provinces,Thailand, 2007–2014

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 7 of 12

pneumoniae, and H. influenzae were common causesof BSIs among 1–4 year olds: S. pneumoniaeaccounted for 13% and H. influenzae 7.6%. Serotypesincluded in the 13-valent pneumococcal conjugatevaccine (PCV13) accounted for 73% of all S. pneumo-niae cases (159 of 218 with serotype available) and92% among children < 5 years old (33 of 36 with sero-type available). Vaccine preventable H. influenzaeserotype B (Hib) accounted for 42% (17 of 40 withserotype available) of all H. influenzae cases and 52%among children < 5 years old. No cases of Neisseriameningitidis were found.Acinetobacter spp., P. aeruginosa, and Enterococcus

faecalis were more common among HO cases comparedto CO cases, while E. coli, B. pseudomallei, and S. pneu-moniae were less common among HO cases than COcases (Fig. 4). Of the B. pseudomallei cases that met theHO definitions, 84% (n = 59) represented the first bloodculture performed during the hospital stay. Twenty ofthe fifty-nine first cultures (34%) were taken on HospitalDay 3 just missing the CO cut-point, an additional 24%

(14/59) were taken on Hospital Day 4 and 86% (51/59)were taken within the first week of hospitalization.ESBL production among E. coli and K. pneumoniae iso-

lates, and MDR among Acinetobacter spp. isolates, wascommon (> 25% of isolates). CO ESBL-producing E. coliand MDR Acinetobacter spp. became more common overtime. The proportions of ESBL producing E. coli and K.pneumoniae, MRSA, and MDR Acinetobacter were 2–3times higher among HO cases compared to CO cases(Table 2). Neither vancomycin resistant S. aureus (VRSA),nor penicillin-resistant S. pneumoniae were observed.Eight cases of CRE were identified between 2011 and2014; six of these cases were also ESBL-producers.

DiscussionPopulation-based surveillance capturing > 130,000 bloodcultures over 8 years documented an overall CO BSI in-cidence of 110 cases per 100,000 population, substan-tially higher than previous reports from Thailand andthe region. This disease incidence is comparable to thatof injuries due to road traffic accidents (86 cases per

Fig. 4 Ten most commonly observed hospital-onset (HO) BSI pathogens by age in Sa Kaeo and Nakhon Phanom provinces

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 8 of 12

100,000 population) and all-site cancers (~ 150 cases per100,000 population) according to WHO Global DiseaseBurden estimates for SE Asia [21].Our 2007–2010 Nakhon Phanom CO BSI incidence

estimates are considerably higher than estimates from 10provincial hospitals in northeast Thailand reported byKanoksil et al.: 91 vs 32.9 (2007); 91 vs 34.6 (2008); 111vs 38.2 (2009); and 118 vs 31.1 (2010) [22]. Our com-paratively higher estimates likely result from threemethodological differences. First, the NESDB NakhonPhanom population estimates used in our incidence cal-culations are smaller than the estimates used by Kanok-sil et al. [23]. Differing population estimates cannotentirely account for this disparity as our estimates arestill significantly higher when we use Department ofProvincial Administration population estimates (data notshown). Second, we included district level hospitals.Kanoksil et al. only captured cases from provincial levelhospitals with the rationale that severely ill patients aretransferred to provincial hospitals, yet we found that44% of pathogen positive cases were collected at districthospitals in Nakhon Phanom province. When ourNakhon Phanom CO BSI estimates were limited to the

provincial level hospital, they ranged between 38.8 cases(2007) and 51.1 (2014), which is comparable to Kanoksilet al.’s estimate for Nakhom Phanom province only: 57.8(2010). Third, our population-based surveillance systemincluded laboratory capacity strengthening activities, re-imbursement for cultures for suspected pneumonia casesand children < 5 years old, and on-going training fornurses and medical technologists, which may have in-creased frequency of blood culture collection, decreasedcontamination rates, and increased culture sensitivity.Previously published bacteremia incidence estimates forSa Kaeo province are not available. Although our COBSI incidence is higher than previously reported, we al-most certainly underestimated the true burden as weonly captured hospitalized patients, pre-culture anti-biotic use was common which lowered culture yield[24], and blood volume targets were frequently missedfor children < 5 years old.Our observed increases in HO BSI incidence in

Nakhon Phanom are consistent with previous reportsfrom Northeast Thailand [25]. Unfortunately, our inci-dence estimates are not directly comparable with thosefrom Hongsuwan et al. as we report cases per 1000

Table 2 Antimicrobial resistance and temporal trends among community-onset (CO) and hospital-onset (HO) BSI pathogens in SaKaeo and Nakhon Phanom provinces, Thailand, 2007-2014

CO HO

Antimicrobial resistance issue % Resistanta

(n/N tested)Temporal Trends % Resistanta

(n/N tested)Temporal Trends

Extended-spectrum beta-lactamase (ESBL) producing E.colib

27% (763/3047)

Increased from 20% in 2008-2010 to 28% in2011-2014

51% (105/203)

No clear trend, range:38% in 2010 to 69% in 2011

ESBL producing K. pneumoniae 23% (213/912)

No clear trend, range:19% in 2014 to 30% in 2012

55% (68/123) No clear trend, range:23% in 2009 to 75% in 2013

Carbapenem-resistantEnterobacteriaceae (CRE)

5 cases No cases before 2011. 5 cases total : 2011 (1),2013 (1), 2014 (3)

3 cases No cases before 2012. 3 casestotal: 2012 (2), 2014 (1).

Methicillin-resistant S. aureus(MRSA)c

7% (55/744) No clear trend 20% (22/111) No clear trend

Vancomycin-resistant S. aureus(VRSA)d

0 0% 0 0%

Penicillin-resistant S.pneumoniaee

0 0% 0 0%

Carbapenem-resistantAcinetobacter spp.

34% (84/242) Increased from 16% in 2007 to >40% in 2008 andthen decreased to 23% in 2014

69% (101/146)

No clear trend, range: 91% in2010 to 46% in 2012

MDR Acinetobacter spp.f 26% (56/219) Increased from <10% in 2007 to >45% in 2010;no clear trend from 2011 (35%) to 2014 (23%)

70% (97/138) Decrease of borderlinesignificance (p<0.10) from 2008to 2014

aCalculated among isolates testedbData available for 2008-2014. ESBL-producing criteria: ≥5 mm increase in the zone of growth inhibition of ceftazidime/clavulanic acid combination disc comparedto ceftazidime or cefotaxime discs alone. If confirmatory testing was not available, ESBL screening test results were used: zone of inhibition for ceftazidime ≤22mm or cefotaxime ≤27 mmcMRSA criteria: prior to 2007, oxacillin disk diffusion with a zone of inhibition <10mm; from 2007, cefoxitin disk diffusion with a zone of inhibition <21 mmdVRSA criteria: MRSA isolates tested for vancomycin minimum inhibitory concentration (MIC) ≤2 ug/ml as determined by E-test (bioMérieux, U.S.A.)ePenicillin resistance criteria: oxacillin disc diffusion zone of inhibition <20 mm with confirmation by a penicillin MIC ≥8 μg/mL by E-testfMDR Acinetobacter criteria: resistant to 3 or more drug classes: aminoglycosides, cephalosporins, fluoroquinolones, carbapenemsgCRE criteria: resistant to imipenem, meropenem, doripenem, or ertapenem by disc diffusion (zone of inhibition <20 mm)

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 9 of 12

hospitalizations, while they reported cases per 1000 hos-pital days. We cannot fully explain the notably higher HOBSI incidence rates in Sa Kaeo province compared toNakhon Phanom, but it may be attributable to differencesin physician practices which is consistent with our obser-vation that hospital stays are longer in Sa Kaeo provincecompared to Nakhon Phanom and blood cultures aremore likely to be performed on patients who have beenhospitalized for > 2 days in Sa Kaeo compared to NakhonPhanom (data not shown). The increasing HO BSI inci-dence in Nakhon Phanom, together with high HO BSI in-cidence rates in Sa Kaeo province, points to a substantial,and potentially growing, HO BSI burden in ruralThailand.With the notable exceptions of B. pseudomallei and Sal-

monella non-typhi spp., our most commonly identifiedCO pathogens, E. coli, K. pneumoniae, S. aureus, and S.pneumoniae, are well-recognized as leading causes of BSIworldwide [26] and in the Asia region [27]. The higherfrequency of B. pseudomallei in our setting was expectedas melioidosis is endemic in these areas of Thailand [6,28], but virtually absent from the study settings includedin Laupland’s review (e.g. Sweden, U.S.A., Denmark,Finland, and England). Similarly, non-typhoidal Salmon-ella spp. were uncommon in these high income countries,

but a documented, common BSI pathogen in ruralThailand [22], Laos [29], and the region [8].We observed substantial increases in CO ESBL pro-

ducing E. coli and MDR Acinetobacter spp. infectionsand emergence of CRE. Previously, we described aMDR prevalence of 31% for healthcare-associated(HCA) Acinetobacter bacteremia cases compared to24% for non-HCA cases for 2005–2008 [30].Healthcare-associated risk factor data needed to clas-sify CO cases HCA vs. non-HCA were not availablefor the current analysis; however MDR Acinetobacterprevalence has clearly increased over time, represent-ing 45–50% of Acinetobacter bacteremia cases from2007 to 2014.This is a laboratory-based surveillance system, supple-

mented with limited clinical data, which restricts ourability to describe several important patient characteris-tics, including outcome. We cannot distinguish primarybloodstream infections from BSI secondary to a focal in-fection, and in some cases, pathogen vs. contaminant.Likewise, we could only crudely classify cases by com-munity- and hospital-onset of infection which is not al-ways a true reflection of the timing of symptom onsetand ultimately, whether the pathogen was acquired inthe community, in the hospital, or another healthcare

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 10 of 12

related setting. We know that a minimum of 11.7%(1176/10,007) of CO cases were hospitalized within theprevious 30 days because these cases have another bloodculture within this time period in our surveillance sys-tem, which only captures blood cultures from hospital-ized patients. However, this underestimates priorhospitalization since we would only know of hospitaliza-tions that included a blood culture. Moreover, we haveno information about other health care exposures (e.g.dialysis or other out-patient treatments). As a more spe-cific example of this limitation, it is likely that many ofthe 70 B. pseudomallei cases that we reported as HO,were actually community-acquired cases that were diag-nosed based on blood cultures taken > 2 days after hos-pital admission.Despite these limitations, this surveillance system has

proved useful in characterizing the public health threatof BSI on many levels: from improving clinical care todeepening regional understanding of BSI epidemiology.By integrating BSI surveillance into routine clinical la-boratory systems, our surveillance provided treating phy-sicians and local partners with accurate, rapid pathogenidentification and antimicrobial susceptibility profiles,which supported appropriate antimicrobial therapy use.Our clinical partners reiterated this message consistentlyand observed improvements in patient outcomes forB. psuedomallei cases after implementation of the au-tomated blood culturing systems support their im-pressions [28, 31]. BSI surveillance has facilitatedtimely responses to public health threats [32, 33],enabled evaluation of new diagnostic tests [34, 35]and incidence and trends in antimicrobial resistance[36–38], and raised awareness of regional emerginginfectious diseases, such as community-associatedAcinetobacter bacteremia, S. suis bacteremia, andbacteremic melioidosis in a part of Thailand not trad-itionally considered highly endemic for that disease[28, 30, 39]. BSI surveillance has contributed to inter-national reports and national health policy evaluations[40, 41]. We expect these data will also contribute tofuture evaluations of key policy issues such asconsideration of PCV and the recent decision to addHib vaccine to Thailand’s Expanded Program onImmunization and strategies to prevent further spreadof anitmicrobial resistant pathogen.Official Thai population projections indicate that be-

tween 2010 and 2030 the number of persons 50 yearsand older will grow by nearly 40% with the largest popu-lation growth occurring in the oldest age-groups [20,19]. Given the high BSI incidence among persons 50years and older, public health officials can expect corre-sponding increases in the BSI disease burden. Inaddition to the costs of acute care, public health officialscan also expect increased costs for long-term care as BSI

survivors suffer from cognitive impairment and func-tional disabilities [42].Improvements in blood culturing systems, BSI surveil-

lance, and other laboratory strengthening efforts areneeded to promptly identify, characterize, and controlthe increasing threat of BSI outbreaks at their source,prevent further emergence of antimicrobial resistance,and enhance global health security. In our experience,these systems are sustainable on a local level. US CDCsupported BSI surveillance stopped in Sa Kaeo provinceas of 2015; however, the Sa Kaeo Provincial Health Of-fice continues automated blood culturing, together withantimicrobial susceptibility testing and infection controlmeasures. Nakhon Phanom provincial health authoritieshave also assumed responsibility for supporting clinicallyindicated blood cultures. In addition to improved bloodculturing systems, population-based BSI surveillancesystems have been shown to be cost-effective in low re-source settings [43] and should be expanded in otherparts of SE Asia. Our finding demonstrate that BSI sur-veillance systems should include all health facilities per-forming blood cultures and incorporate laboratorystrengthening activities to ensure accurate measurementof the full BSI disease burden, facilitate rapid detectionand control of outbreaks at their source, and thereby en-hance global health security.

ConclusionsHigh CO BSI incidence in rural Thailand demonstratesthe need for on-going surveillance with laboratorystrengthening to improve clinical care and prevent fur-ther emergence of antimicrobial resistance.

Additional file

Additional file 1: Figure S1. Community-onset (CO) BSIs caused byGram-negative pathogens by month in Sa Kaeo and Nakhon Phanomprovinces, Thailand, 2007–2014. (PDF 156 kb)

Abbreviations°C: Celsius; BSI: Bloodstream infection; CI: Confidence Intervals;CO: Community-Onset (CO; CRE: Carbapenem-Resistant Enterobacteriaceae;EQA: External Quality Assurance; ESBL: Extended Spectrum Beta-Lactamase;FA: FAN Aerobic (FA); HO: Hospital-Onset (HO); ICUs: Intensive Care Units;MB: Mycobacteria Blood (MB); MDR: Multi-Drug Resistant; MRSA: Methicillin-Resistant S. aureus; NESDB: National Economic and Social DevelopmentBoard; NIH: National Institute of Health; NP: Nakhon Phanom (NP); PCR: PCR;PF: Pediatric FAN (PF); PneumoADIP: Pneumococcal Vaccines AcceleratedDevelopment and Introduction Plan; SAS: Statistical Analysis Software; SK: SaKaeo; US CDC: United States Center for Disease Control and Prevention

AcknowledgmentsFor their contributions to this work, we thank Patranuch Sapchookul, PattrapornKlanjatturat, Barameht Piralam, Sathapana Naorat, Ying Lu, PongpunSawatwong, Duangkamon Siludjai, Apiwat Lapamnouysup, (GlobalDisease Detection Regional Center, Thailand Ministry of Public Health-U.S.Centers for Disease Control and Prevention [CDC] Collaboration). We alsothank our colleagues, collaborators and partners at the Nakhon Phanomand Sa Kaeo Provincial Health Offices.

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 11 of 12

FundingSupport for this project was provided by the CDC Foundation, thePneumococcal Vaccines Accelerated Development and Introduction Plan(PneumoADIP) (funded by the GAVI Alliance based at the Johns HopkinsBloomberg School of Public Health), and the Global Disease Detectionprogram of the U.S. Centers for Disease Control and Prevention. Publicationcosts were provided by the Global Disease Detection program.

Availability of data and materialsThe data that support the findings of this study are from Thailand Ministry ofPublic Health - US CDC Collaboration, but restrictions apply to the availabilityof these data, and so are not publicly available. Data are however availablefrom the authors upon reasonable request and with permission of the NakhonPhanom and Sa Kaeo Provincial Health Offices.

About this supplementThis article has been published as part of BMC Public Health Volume 19Supplement 3, 2019: 10th anniversary of the Centers for Disease Control andPrevention - Global Disease Detection program. The full contents of thesupplement are available online at https://bmcpublichealth.biomedcentral.com/articles/supplements/volume-19-supplement-3.

Authors’ contributionsJR –contributed to acquisition of data, analysis and interpretation of data,and drafting and revising the article. PJ, SM, OS, AK, TA, PS, ST, SK, SY, WP,AK, TW, HB, and CG contributed to conception and design, acquisitionof data, analysis and interpretation of data, and revising the article. Allauthors read and approved the final manuscript.

Ethics approval and consent to participateThe CDC Human Subjects Review Office reviewed this protocol and judgedthat this study constituted routine public health activities and therefore didnot involve human subject research (CGH Determination and Approvalnumber 2014–273).

Consent for publicationNot applicable.

Competing interestsThe authors declare that they have no competing interests.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims in publishedmaps and institutional affiliations.

Author details1Global Disease Detection Center, Thailand Ministry of Public Health (MOPH)– United States Centers for Disease Control and Prevention (CDC)Collaboration, Nonthaburi, Thailand. 2Division of Global Health Protection,Center for Global Health, CDC, Atlanta, GA, USA. 3Nakhon Phanom GeneralHospital, Nakhon Phanom, Thailand. 4Sa Kaeo Crown Prince Hospital, SaKaeo, Thailand. 5Department of Medical Sciences, National Institute of Health,Ministry of Public Health, Nonthaburi, Thailand. 6Faculty of Public Health,Kasetsart University Chalermphrakiat, Sakon Nakhon Province, Thailand.

Published: 10 May 2019

References1. Laupland KB, Gregson DB, Flemons W, Hawkins D, Ross T, Church DL.

Burden of community-onset bloodstream infection: a population-basedassessment. Epidemiol Infect. 2007;135:1037–42.

2. Goto M, Al-Hasan MN. Overall burden of bloodstream infection andnosocomial bloodstream infection in North America and Europe. ClinMicrobiol Infect. 2013;19:501–9.

3. Chayakulkeeree M, Junsriwong P, Keerasuntonpong A, Tribuddharat C,Thamlikitkul V. Epidemiology of extended-spectrum beta-lactamasesproducing gram-negative bacilli at Siriraj hospital, Thailand, 2003. SoutheastAsian J Trop Med Public Health. 2005;36(6):1503–9.

4. Mekviwattanawong S, Srifuengfung S, Chokepaibulkit K, Lohsiriwat D,Thamlikikul V. Epidemiology of Staphylococcus aureus infections and theprevalence of infection caused by community-acquired methicillin-resistant

Staphylococcus aureus in hospitalized patients at Siriraj hospital. J Med AssocThail. 2006;89(Suppl 5):S106–S17.

5. Nickerson EK, West TE, Day NP, Peacock SJ. Staphylococcus aureus diseaseand drug resistance in rural Asia. Lancet Infect Dis. 2009;9:130–5.

6. Limmathurotsakul D, Wongratanacheewin S, Teerawattanasook N, Wongsuvan G,Chaisuksant S. Increasing incidence of human melioidosis in Northeast Thailand.Am J Trop Med Hyg. 2010;82:1113–7. https://doi.org/10.4269/ajtmh.2010.10-0038.

7. Luvsansharav U-O, Hirai I, Niki M, et al. Analysis of risk factors for a highprevalence of extended-spectrum b-lactamase-producingEnterobacteriaceae in asymptomatic individuals in rural Thailand. J MedMicrobiol. 2011;60:619–24. https://doi.org/10.1099/jmm.0.26955-0.

8. Deen J, von Seidlein L, Andersen F, et al. Community-acquired bacterialbloodstream infections in developing countries in south and SoutheastAsia: a systematic review. Lancet Infect Dis. 2012;12:480–7.

9. Jorakate P, Higdon M, Kaewpan A, et al. Contribution of the BacT/Alert MBmycobacterium bottle to bloodstream infection Sureveillance in Thailand:added yield for Burkholderia pseudomallei. J Clin Microbiol. 2015;53:910–4.

10. Jorgensen JH, Pfaller MA, Carroll KC, et al. Manual of clinical microbiology.11th ed. Washington, DC: ASM Press; 2015. https://doi.org/10.1128/9781555817381.

11. Clinical and Laboratory Standards Institute (CLSI). Performance Standards forAntimicrobial Susceptibility Testing; Twenty-Third Informational Supplement.Wayne, PA: CLSI; 2013.

12. Giriyapur RS, Nandihal NW, Krishna B, Patil AB, Chandrasekhar M.Comparison of disc diffusion methods for the detection of extended-Spectrum Beta lactamase-producing Enterobacteriaceae. J Lab Physicians.2011;3(1):33–6.

13. Morrissey I, Bouchillon SK, Hackel M, et al. Evaluation of the clinical andlaboratory standards institute phenotypic confirmatory test to detect thepresence of extended-spectrum b-lactamases from 4005 Escherichia coli,Klebsiella oxytoca, Klebsiella pneumoniae and Proteus mirabilis isolates. J MedMicrobiol. 2014;63:556–61.

14. Siegel JD, Rhinehart E, Jackson M, Linda C. Healthcare Infection ControlPractices Advisory Committee. Management of multidrug-resistantorganisms in healthcare settings (2006). Available at: https://www.cdc.gov/infectioncontrol/guidelines/mdro/.

15. Pai R, Gertz RE, Beall B. Sequential multiplex PCR approach for determingcapsular serotypes of Streptococcus pneumoniae isolates. J Clin Microbiol.2006;44(1):124–31. https://doi.org/10.1128/JCM.44.1.119-123.2006.

16. Falla TJ, Crook DW, Brophy LN, Maskell D, Kroll JS, Moxon ER. PCR for capsulartyping of Haemophilus influenzae. J Clinical Microbiol. 1994;32(10):2382–6.

17. Hall KK, Lyman JA. Updated review of blood culture contaminates. ClinMicrobiol Rev. 2006:788–802. https://doi.org/10.1128/CMR.00062-05.

18. Raymund D, Yi M, Ruth B, et al. National burden of invasive methicillin-resistant Staphylococcus aureus infections, United States, 2011. JAMA InternMed. 2013;173(21):1970–8.

19. Thailand NESDB. Population projections for Thailand 2010–2040. Availableat: http://social.nesdb.go.th/social/Portals/0/Documents/%E0%B8%81%E0%B8%B2%E0%B8%A3%E0%B8%84%E0%B8%B2%E0%B8%94%E0%B8%9B%E0%B8%A3%E0%B8%B0%E0%B8%A1%E0%B8%B2%E0%B8%93%20e-book.pdf.

20. Thailand NESDB. Population projections of Thailand 2000–2030. Available at:http://social.nesdb.go.th/social/Portals/0/Documents/pop_34.zip.

21. World Health Organization. The global burden of disease: 2004 update.Geneva: WHO Press; 2008. Available at: https://www.who.int/healthinfo/global_burden_disease/2004_report_update/en/

22. Kanoksil M, Jatapai A, Peacock SJ, Limmathurotsakul D. Epidemiology,microbiology and mortality associated with community-acquiredbacteremia in Northeast Thailand: a multicenter surveillance study. PLoSOne. 2013;8(1):e54714. https://doi.org/10.1371/journal.pone.0054714.

23. Department of Provincial Administration. Thailand Demographic. Availableat: http://stat.bora.dopa.go.th/hpstat9/people2.htm.

24. Rhodes J, Hyder JA, Peruski LF, Fisher C, Jorakate P. Anitbiotic use inThailand: quantifying impact on blood culture yield and estimates ofpneumococcal bacteremia incidence. Am J Trop Med Hyg. 2010;83(2):301–6.https://doi.org/10.4269/ajtmh.2010.09-0584.

25. Hongsuwan M, Srisamang P, Kanoksil M, Luangasanatip N. Increasingincidence of hospital-acquired and healthcare-associated bacteremia inNortheast Thailand: a multicenter surveillance study. PLoS One. 2014;9(10):e109324 Published 2014 Oct 13. doi:10.1371/journal.pone.0109324.

26. Laupland K. Incidence of bloodstream infection: a review of population-based studies. Clin Microbiol Infect. 2013;19:492–500.

Rhodes et al. BMC Public Health 2019, 19(Suppl 3):521 Page 12 of 12

27. Moon HW, Ko YJ, Park S, Hur M, Yun YM. Analysis of community- andhospital-acquired bacteraemia during a recent 5-year period. J MedMicrobiol. 2014;63:421.

28. Bhengsri S, Baggett HC, Jorakate P, et al. Incidence of Bacteremic Meliodosisin eastern and northeastern Thailand. Am J Trop Med Hyg. 2011;85(1):117–20.https://doi.org/10.4269/ajtmh.2011.11-0070.

29. Phetsouvanh R, Phongnamy S, Soukaloun D, et al. Causes of community-acquired bacteremia and patterns of antimicrobial resistance in Vientiane,Laos. Am J Trop Med Hyg. 2006;75(5):978–85.

30. Porter KA, Rhodes J, Dejsirilert S, et al. Acinetobacter bacteraemia inThailand: evidence for infections outside the hospital setting. EpidemiolInfect. 2014;142(6):1317–27.

31. Jatapai A, Gregory CJ, Thamthitiwat S, et al. Hosptitalized bacteremic meliodosisin rural Thailand: 2009-2013. Am J Trop Med Hyg. 2018;98(6):1585–1591.

32. Thamthitiwat S, Marin N, Baggett HC, et al. Mycobacterium bovis (BacilleCalmette-Guerin) bacteremia in immunocompetent neonates followingvaccination. Vaccine. 2011;29:1727–30.

33. Thamthitiwat S, Chuxnum T, Baggett HC, et al. Human Brucella Abortusinfection in Thailand: A Report of the First Two Cases. In: Am J Trop MedHyg 56th Annual Meeting Abstract book. Philadelphia, PA, USA: http://www.astmh.org/annual-meeting/past-meetings, 2007:260.

34. Lindsley MD, Mekna N, Baggett HC, Surinthong Y, Autthateichai R.Evaluation of a newly developed lateral flow immunoassay for the diagnosisof Cryptococcosis. CID. 2011;53(15 August):321–5.

35. Baggett HC, Rhodes J, Dejsirilert S, Salika P, Wansom T. Pneumococcalantigen testing of blood culture broth to enhance the detection ofStreptococcus pneumoniae bacteremia. Eur J Clin Microbiol Infect Dis. 2012;31:753 https://doi.org/10.1007/s10096-011-1370-3.

36. Whistler T, Sapchookul P, McCormick DW, et al. Epidemiology andantimicrobial resistance of invasive non-typhoidal salmonellosis in ruralThailand from 2006-2014. PLoS Negl Trop Dis. 2018;12(8):e0006718. https://doi.org/10.1371/journal.pntd.0006718.

37. Jaganath D, Jorakate P, Makprasert S, et al. Staphylococcus aureusbacteremia incidence and methicillin resistance in rural Thailand, 2006–2014. Am J Trop Med Hyg. 2018;99(1):155–63.

38. Sawatwong P, Sapchookul P, Whistler T, et al. High burden of extended-spectrum β-lactamase producing Escherichia coli and Klebsiellapneumoniae bacteremia in older adults: a seven - year study in two ruralThai provinces. Am J Trop Med Hyg. 2019;100(4):943–51. https://doi.org/10.4269/ajtmh.18-0394.

39. Praphasiri P, Owusu JT, Thammathitiwat S, et al. Streptococcus suis infectionin hospitalized patients, Nakhon Phanom Province, Thailand. Emerg InfectDis. 2015;21:345–8. https://doi.org/10.3201/eid2102.140961.

40. Johnson HL, Deloria-Knoll M, Levine OS, Stoszek SK, Hance LF. Systematicevaluation of serotypes causing invasive pneumococcal disease amongchildren under five: the pneumococcal global serotype project. PLoS Med.2010;7(10):e1000348. https://doi.org/10.1371/journal.pmed.1000348.

41. Kulpeng W, Leelahavarong P, Rattanavipapong W, et al. Cost-utility analysisof 10- and 13-valent pneumococcal conjugate vaccines: protection at whatprice in the Thai context? Vaccine. 2013;31:2839–47.

42. Iwashyna TJ, Ely EW, Smith DM, Langa KM. Long-term cognitive impairmentand functional disabilty among survivors of severe sepsis. JAMA. 2010;304(16):1787–94. https://doi.org/10.1001/jama.2010.1553.

43. Penno EC, Baird S, Crump JA. Cost-effectiveness of surveillance forbloodstream infections for sepsis management in low-resource settings. AmJ Trop Med Hyg. 2015;93(4):850–60. https://doi.org/10.4269/ajtmh.15-0083.